Based on the documentation, the function tf.keras.backend.ctc_decode is supposed to return a Tuple. Its first field contains the best path (let's assume we use greedy search), whereas the second one contains its log probability.
Is this probability actually the accuracy of the prediction?
If not how am I supposed to calculate it?
I've tried on some test images and this was my output:
True value: test0, prediction: test0, acc: 1.841524362564087
True value: test1, prediction: test1, acc: 0.9661365151405334
True value: test2, prediction: test2, acc: 1.0634151697158813
True value: test3, prediction: test3, acc: 2.471940755844116
True value: test4, prediction: test4, acc: 1.4866207838058472
True value: test5, prediction: test5, acc: 0.7630811333656311
True value: test6, prediction: test6, acc: 0.35642576217651367
True value: test7, prediction: test7, acc: 1.5693446397781372
True value: test8, prediction: test8, acc: 0.9700028896331787
True value: test9, prediction: test9, acc: 1.4783780574798584
During the training part, the final CTC loss was around 0.1 and the prediction is always correct. However what I think it's the probability seems not to be what I expect. They looks like completely random numbers, even grater than 1 or 2! What am I doing wrong?